Evaluation of Rainfall Erosivity Factor Estimation Using Machine and Deep Learning Models

Rainfall erosivity factor (R-factor) is one of the Universal Soil Loss Equation (USLE) input parameters that account for impacts rainfall intensity in estimating soil loss. Although many studies have calculated R-factor using various empirical methods or USLE method, these are time-consuming and require specialized knowledge user. The purpose this study to develop machine learning models predict faster more accurately than previous methods. For this, 1-min interval data improved accuracy target value. First, monthly R-factors were calculation method identify characteristics rainfall-runoff induced erosion. In turn, developed at 50 sites Korea as values. algorithms used Decision Tree, K-Nearest Neighbors, Multilayer Perceptron, Random Forest, Gradient Boosting, eXtreme Boost, Deep Neural Network. As a result validation with 20% randomly selected data, Network (DNN), among seven models, showed greatest prediction results. DNN was tested six demonstrate trained model performance Nash–Sutcliffe Efficiency (NSE) coefficient determination (R2) 0.87. This means our findings show can be efficiently estimate desired site much less effort time total precipitation, maximum daily hourly precipitation data. It will not only calculate erosion risk but also establish conservation plans areas disasters by calculating factors.